Post hoc support vector machine learning for impedimetric biosensors based on weak protein-ligand interactions

Area of Science:

• Biosensing and Nanotechnology
• Machine Learning Applications
• Analytical Chemistry

Background:

• Impedimetric biosensors face challenges in detecting small molecules due to weak interactions, especially in complex samples.
• Accurate analysis of impedance spectra from protein-ligand binding sensors is difficult, requiring computationally intensive methods.
• There is a need for efficient, real-time analytical tools for impedimetric biosensor data.

Purpose of the Study:

• To introduce a simple, open-source support vector machine (SVM) algorithm for analyzing impedimetric data.
• To demonstrate the SVM tool's effectiveness across different protein-based biosensor applications.
• To validate the SVM tool's performance against traditional equivalent circuit analysis.

Main Methods:

• Developed and applied an open-source SVM machine learning algorithm using Python and scikit-learn.
• Utilized Jupyter Notebook for data analysis and deployment, ensuring accessibility.
• Tested the algorithm on two distinct protein-based biosensors, including a mobile phone-based acetone sensor.

Main Results:

• The SVM algorithm performed comparably to or better than equivalent circuit analysis for weak/transient interactions.
• Demonstrated successful application in analyzing acetone, a biomarker for diabetic ketoacidosis.
• The tool exhibits low computational requirements, suitable for mobile biosensing platforms.

Conclusions:

• The open-source SVM algorithm provides an effective and computationally inexpensive alternative for analyzing impedimetric biosensor data.
• The tool facilitates rapid detection and can be integrated into mobile acquisition systems.
• This approach supports the development of nanobiosensors for planetary health monitoring using mobile technology.